Governance
Governance is what makes TapPass different from a proxy. Every action an agent takes is decided by the governance kernel, facts in, policy verdict out, and every decision lands in a signed, hash-chained audit trail.
The moving parts
Section titled “The moving parts”1. The kernel, Every call is evaluated against the agent’s active policy; the decision is recorded. See How TapPass works.
2. Policy, versioned rules, assigned at org → project → team → agent scope, rolled out through draft → shadow → active. See Authoring policies and the rule reference.
3. Governance Flags, per-request intent your agent code sets via one header.
4. Compliance Evidence, the audit trail as GDPR, EU AI Act, and SOC 2 evidence: integrity verification, replay, DSAR export, retention.
What happens per call
Section titled “What happens per call”request ─▶ authenticate ─▶ evaluate policy ─▶ apply obligations ─▶ record ─▶ provider/tool │response ◀─ audit ◀─ response govern ◀────────────────────────────────────────┘- Facts are collected on demand, Only the checks your policy needs are evaluated.
- The verdict is deterministic, replayable months later.
- Enforcement is fail-closed, missing policy, unreadable bundle, or a failing detector blocks the call rather than waving it through.
- Blocks are explainable, every decision names the rule, the evidence, and the policy version that decided it.
What you control, from outside in
Section titled “What you control, from outside in”| Layer | Who controls it | Takes effect |
|---|---|---|
| Policy rules & assignments | Platform/security team | On publish (atomic) |
| Shadow → enforce rollout | Platform team | Per version |
| Flags (per-request intent) | Agent developers | Per call |
| Compliance exports & retention | Compliance team | On demand |
Where to go next
Section titled “Where to go next”- New to the model? Read the Concepts section in order.
- Ready to write policy? Start with Authoring policies.
- Need to prove it? Go to Compliance Evidence.